As originally published in the SMTA Preedings.

BGA RE-BALLING FROM PB-FREE TO SN-PB METALLURGY

S. J. Meschter, S.A. McKeown, R. Feathers and E. Arseneau
BAE Systems
Johnson City, NY, USA
stephan.j.meschter@baesystems.com

ABSTRACT ΔT = Change in temperature, °C

As a result of a global movement away from using Lead Tg = Glass transition temperature
(Pb) in electronic assemblies, component manufacturers are W = Width
almost exclusively providing lead-free parts to satisfy the
high volume consumer markets. Unfortunately, relatively BACKGROUND
little is known about the performance of lead-free solders in Recent restriction of hazardous substances (RoHS)
harsh vibration and shock environments. These concerns are legislation from the European Union [1] as well as current
amplified because the consumer industry is currently environmentally friendly market trends [2] have resulted in
evaluating another generation of lead-free solder alloys in the elimination of lead (Pb) from many electrical and
an effort to improve reliability. Tin whiskers not electronic assemblies. Historically, electronic components
withstanding, nearly all the current lead-free electronic have been manufactured using a tin-lead solder (typically,
piece-part termination finishes are compatible with tin-lead Sn63Pb37). The reliability of the tin-lead solders has been
assembly solder with the exception of ball grid arrays. well documented and proven. The change from tin-lead to
Reprocessing lead-free BGAs with tin-lead ball metallurgy lead-free materials may be tolerable for many consumer
is one means of mitigating the risk of lead-free solder electronic devices having a short life time and minor
material failure modes such as tin whiskers, high cycle consequences of failure. However, lead-free use in long life
fatigue, printed circuit board pad cratering, and intermetallic high reliability systems having high consequences of failure
fracture. In addition, because qualification of a metallurgy that are used in the back bone of the defense, transportation,
change in a high reliability application can take years, BGA communication, and medical infrastructure needs to be fully
reballing allows original equipment manufacturers to understood.
maintain the certification and qualification status on existing
configurations while managing the on-going lead-free alloy The introduction of lead-free solder has brought on
changes occurring on BGAs. In the present work, an reliability concerns in high reliability applications having
assessment of the mechanical integrity of four different long service life and significant repair activity throughout
commercially available BGAs was evaluated after reballing the product’s life cycle [3][4]. In addition to the tin whisker
using visual inspection, cross-section evaluation, scanning risks accompanying lead-free materials, designers are
acoustic microscopy, moiré interferometry, ball shear, ball challenged by the higher processing temperatures, reduced
pull and assembly level thermal cycling. vibration/shock performance and increased occurrences of
brittle fractures of intermetallics and printed circuit board
Key Words: Lead-free, Ball Grid Array, Solder, Reliability, pads. Furthermore it may be undesirable to utilize lead-free
Electronics, Reballing bismuth-bearing solder alloys in high reliability repair
depots because inadvertent mixing of tin-lead and bismuth
Nomenclature and abbreviations: bearing lead-free alloys can result in moderate to dramatic
BGA = Ball grid array reductions in reliability [5][6].
CTE = Coefficient of thermal expansion
CSAM = C-mode scanning acoustic microscope High reliability designs have a long history of using
Dia. = Diameter mainstream consumer electronics and the DoD recognizes
Dims. = Dimensions the advantages of dual use and commercial off the shelf
ENIG = Electroless nickel immersion gold (COTS) items. Sometimes it is possible to use the items
L = Length directly, in other instances it is necessary to select parts
ΔL = Change in length of the specimen based on certain performance parameters, and sometimes it
Ni = Nickel is necessary to alter items in order to meet application
Nf = Number of fringes obtained requirements. In the present situation where lead-free ball
P = Phosphorous grid arrays are unsuitable for use, it is possible to alter them
Pb = Lead by removing the lead-free balls and replacing them with tin-
PWB = Printed wiring board lead balls [7] [8]. For this investigation, when reballing
SAM = Scanning acoustic microscope suppliers were approached to reball from lead-free to tin-
SMD = Solder mask defined lead, several indicated that they already had considerable
Sn = Tin experience reballing tin-lead BGAs to lead-free during the
T = Thickness early stages of the RoHS transition.
 As originally published in the SMTA Preedings.

During the BGA reballing process, the lead-free balls are electroless nickel (e.g. NiP) layer over the copper exhibit
first removed, then the package is cleaned, the new tin-lead very little pad dissolution from the solder during reballing.
balls are attached and the package is cleaned again. One of Although use of gold over copper conductor under solder
the more challenging processes is ball removal because it mask is not very prevalent, it does substantially increase the
tends to deviate from the standard surface mount reflow likelihood of solder mask separation which can result in
processes that BGAs are designed to withstand. Two solder tunneling under the solder mask [10]. The last part of
common methods of ball removal are (1) solder wick and the preliminary assessment in Fig. 2 is motivated by the fact
(2) flowing wave [7] [8]. The solder wick method has the that scanning acoustic microscopy (SAM) cannot be used to
advantage of lower overall package temperatures during assess the internal integrity of some BGAs. Among the
processing than the flowing wave ball removal process. BGA features that inhibit acoustic microscopy are (1)
However, solder wick has the disadvantage of being a packages with lids and cavities, (2) use of low modulus glob
manual process that has locally higher heating at the top layers over the die, (3) use of low acoustic density layers
individual solder pads. In addition, since solder wick braid within the part and (4) complex BGA interconnect
comes in contact with the BGA during processing, there is structures that attenuate acoustic energy. In the case where
increased risk of solder mask damage. SAM cannot be used, other assessment methods must be
employed.
Once the balls have been removed, and the bottom side of
the package has been visually inspected, the new tin-lead When developing reballing acceptance requirements, it is
balls are then reattached. Ball reattachment also involves important to avoid introducing requirements that exceed the
thermal processing of the part. Often a standard surface original piece part requirements (e.g. warpage requirements
mount technology convection reflow process is used, but after reballing cannot be more stringent than the
some reballing suppliers utilize laser soldering that typically requirements of the original part). An additional item that
results in low overall package heating. As long as the must be considered is that all manufacturing processes (e.g.
package integrity is maintained, the likelihood of electrical dry baking, board soldering or environmental stress
issues is low for most BGAs, especially since most are screening) result in some amount of change to the part. With
digital devices. In one study, extensive testing that included that in mind, some part changes encountered during
base loopback, top loopback, memory, flash, script and reballing may be acceptable and will not impact the
SRAM of devices after reballing showed no failures [8]. reliability in the intended application.

Molding compound regions EVALUATION METHODOLOGY

Die attach adhesive
Die Over die Adjacent to die The BGAs under study are actual functional BGAs obtained
from four different mainstream component manufacturers
(designated M, L, F and X) so that real process variation
from the supply chain would be included in the assessment.
The sample type designation M90 indicates that the part is a
90 ball BGA from supplier M. These BGAs were selected to
encompass a range of package sizes, die sizes, ball
Solder mask thickness BGA Interconnect diameters and ball pitches as shown in Table 1 and Fig. 3.
Three packages have wire bonded die (M90, L256 and
thickness F473) and one is a flip chip (X1148). Two of the BGAs
Solder mask defined (SMD)
were only available with lead-free balls. The L256 was
package ball attach pad diameter
available with either tin-lead or lead-free ball metallurgy
Fig. 1: BGA structure. allowing a unique comparison of various assessment
parameters. The X1148 was only available with tin-lead ball
BGAs have relatively complex internal structures (Fig. 1)
metallurgy and was included to evaluate a tin-lead ball
that can vary considerably from part to part. Since the
replacement/repair process.
reballing processing adds additional heat cycles to the BGA,
the present work seeks to evaluate the mechanical integrity
The evaluations performed and the quantities of parts used
of the package after reballing.
in the present work are summarized in Table 2 and Fig. 4.
In order to ensure a proper match between BGA type and The basic characteristics of the parts were obtained to
reballing process an acceptance assessment is needed. Given facilitate modeling and future similarity analyses.
the broad range of BGA constructions, it is likely that some
BGA Construction and Integrity Evaluation
BGAs will be tolerant to a broad range of reballing methods,
Overall package measurements were made and compared
while others will need more controlled process limits. An
with the datasheet. Photographs were obtained to record
overall BGA reballing acceptance process flow is outlined
overall package details including the package marking.
in Fig. 2. Clearly, experience with a particular BGA
Cross-sectioning in conjunction with optical and scanning
manufacturer and construction type with a particular
electron microscopy and radiographic imaging were used to
reballing process is a discriminator when formulating the
determine the internal construction details needed for
BGA reballing assessment plan. BGAs with a nickel or an
modeling and similarity analysis as shown in Fig. 1. The
 As originally published in the SMTA Preedings.

cross-section also allowed an assessment of the internal A detailed cross-section evaluation was also performed on
molding compound and laminate integrity, the condition of the ball attach pads. The ball attach pad analysis included an
the BGA interconnect board vias and traces, the internal die optical or SEM image of a minimum of three BGA ball
to package connections, as well as the solder mask adhesion attach pads with balls removed showing diameter of BGA
to the BGA interconnect pads and surface copper features. ball attach pad, thickness measurements of the copper and
the nickel layers, a determination of the nickel type
Preliminary BGA reballing assessment (electroless nickel, e.g. NiP alloy, or electrolytic nickel), and
 Is there prior reballing experience with the BGA an assessment of the intermetallic. Since cross-sectioning is
construction and are the materials known? a destructive analysis, different BGAs were assessed before
 Is there prior reballing experience with the BGA and after reballing. The cross-sectioning was complemented
piece part manufacturer? by radiographic imaging to determine the internal features
 Does the BGA have Ni or NiP plated ball of interest.
attachment pads?
 No gold over copper conductors are used on the
solder side of the BGA?
(A) (B) (C)
 Can the BGA be examined with SAM?
M90: (A) Top, (B) Bottom, and (C) radiographic
Yes No
Perform baseline BGA Possible supplemental
acceptance with selected evaluations to standard
reballing process (es) acceptance: (A) (B) (C)
 Acceptance by  Part L256: (A) top, (B) bottom, and (C) radiographic images
similarity Cross-section
OR Radiographic eval.
 Pre and post-reball Multi removal/attaches
visual and acoustic Warpage and/or CTE
microscopy Ball shear and/or pull
 Dimensional Electrical functional (A) (B) (C)
evaluation  Assembly F473: (A) Top, (B) Bottom, and (C) radiographic
 Electrical evaluation Electrical functional
may be at a piece part Thermal cycling
or at a higher assembly Vibration
level Mechanical Shock

Engineering evaluates acceptability of results

Acceptable Not acceptable
(A)
Evaluate alternate reballing
Add part to altered processes, alternate parts,
item drawing repackaging, lead-free solder
assembly, or other options

Fig. 2: Overall BGA reballing process flow

Table 1: BGA types and dimensions being evaluated (mm)
(B) (C)
BGA Pkg dims. Ball Ball Ball Die X1148: (A) Top, (B) Bottom, and (C) radiographic
(LxWxT) dia. pitch extents size
(LxW) Fig. 3: Photomicrograph and radiographic images of the
M90 13x10x0.65 0.45 0.8 11.2 7.53 M90, L256, F473 and X1148 BGAs being evaluated.
x 6.4 x6.79
Acoustic Microscopy and Visual Assessment Methods
L256 17x17x1.25 0.5 1.0 15 x 15 4.28 Scanning acoustic microscopy (SAM) and detailed visual
x3.94 inspections were performed on a sample of parts before and
F473 19x19x1.12 0.4 0.8 17.6 6.23 after reballing. SAM was performed to identify any
x 17.6 x5.68 delamination within the part at various critical intrapackage
X1148 35x35x 2.8 0.6 1.0 33 x 33 19.3 interfaces such as the die-top and molding compound, die
x13.6
 As originally published in the SMTA Preedings.

bottom and die attach adhesive, and the interconnect layer. Furthermore, since reflected sound waves are generated
SAM can also be useful for determining die size. Visual each time a different acoustic density is encountered in the
inspection was used to assess any surface breaking features. interconnect area, reflections are generated from the copper
Often, combinations of C-mode reflected wave acoustic traces, copper plane layers, vias, glass fibers, and solder
imaging (CSAM) and through scan acoustic microscopy can mask layers making it difficult to resolve delaminated
be used to evaluate the internal state of delamination in regions.
electronic parts. In the present work, equipment issues
prevented using a through scan assessment and in one case
cross-sectioning was needed to verify interface integrity. Total: 90 BGAs procured for each BGA type
Scanning acoustic microscopy could not be used to assess
the X1148 BGAs since acoustic waves could not traverse 20 BGAs: Pre-reballing evaluation
the void region under the lid.
Non-destructive
Table 2: Parameters evaluated for each BGA type  12 BGAs: Scanning acoustic microscopy and
Parameter Description visual delamination evaluation *
Destructive
One sample
 1 BGA: Cross-sectioning
Microphotographs of the top, side and
Overall  2 BGA: Warpage
bottom of the package being sure to
photographs  4 BGA: CTE, ball shear, ball pull **
capture all marking.
Radiographic Determine die size and any significant Spare control samples
examination metal structures.  1 BGA: Not to be reballed
BGA ball Assessment of BGA ball diameter
diameter uniformity by reballing supplier. 82 BGAs: Total parts to be reballed
Obtain internal construction details  12 BGAs: From previous scanning acoustic
Cross-section
needed for modeling and similarity microscopy and visual delamination evaluation
of the package
analysis.  72 BGAs: Additional parts to be reballed
Cross-section Cross-section assessment of a typical
of BGA ball ball attach pad before and after re-
attach pad balling. 22 BGAs: Post-reballing part evaluation and spares
BGA Pad Measure amount of BGA pad metal Non-destructive
metallization dissolution due to re-balling.  12 BGAs: Scanning acoustic microscopy and
Pad Evaluate intermetallic voiding, visual delamination evaluation (save as spares)
intermetallic(s) cracking, morphology evaluation Destructive
Performed using cross-sectional Moiré  1 BGA: Cross-section, nickel thickness
Package CTE
interferometry  2 BGAs: Warpage
Shadow Moiré interferometry  4 BGAs: CTE, Ball shear and ball pull **
Package measurements giving package warpage Spare reballed control samples
warpage over the use and soldering  3 BGA: Not assembled
temperatures.
Ball shear High speed ball shear testing was
before and performed according to JEDEC 60 BGAs: Post-reballing assembly evaluation ***
after reballing JESD22-B117A at a velocity of 1m/s.  1 board (20 BGAs): Thermal Cycling (In process)
Ball pull High speed pull testing was performed  1 board (20 BGAs): Vibration (Planned)
before and according to JEDEC JESD22-B115 at  1 board (20 BGAs): Mechanical Shock (Planned)
after reballing a velocity of 100 mm/s.
Ten samples Notes:
Scanning * A minimum of ten BGAs are needed for acceptance
C-Mode and through scan acoustic testing. Some additional parts were included in case
acoustic
microscopy for delamination evaluation anomalies in as-received parts would prevent
microscopy
Visual 40x optical inspection for external evaluation.
inspection cracks. ** ½ of the part is needed for CTE so ½ can be used for
All parts ball shear and/or ball pull
In accordance with J-STD-001, any *** 20 BGAs are needed for each test board assembly.
Visual Each board has 12 functionally monitored BGAs and
visual damage in excess of part
inspection four BGAs for cross-sectioning with four reworked
specification is cause for rejection.
Verification Ball diameter measured for compliance BGAs (removed and replaced).
of ball with original manufacturer’s Fig. 4: Reballing BGA part utilization for each BGA type.
diameter requirements
 As originally published in the SMTA Preedings.

The C-mode scanning transducer head frequency was Two of each BGA type were prepared by mechanically
selected in accordance with IPC/JEDEC J-STD-035 removing the solder balls using a shear tool. The ball-attach
paragraph 4.1.1 to obtain maximum image resolution at the side of each sample was then lightly dusted with a coat of
regions of interest. A transducer ranging in frequency from high temperature white paint to gain the necessary contrast
15 to 50 MHz was used, with corresponding spot sizes from for the measurement technique. All warpage measurements
0.180 to 0.073 mm. The coupling medium used in the were made on the ball-attach side of the components. The
containment tank was distilled water. measurements were taken with the solder ball side facing
up. The thermal profile used to test each sample is shown in
The amount of delamination detected over each area of Fig. 5. A peak temperature of 245 °C was selected, and
interest was totaled by an automated process before and measurements were made at the temperatures indicated by
after reballing. A minimum of ten parts were individually the circles in the profile figure. Of particular interest were
serialized and evaluated for cracking/delamination before the warpages near the maximum service temperatures (100
and after reballing. Parts exhibiting anomalies preventing and 125 °C), the warpage near the tin-lead eutectic melting
measurements before re-balling are allowed to be replaced, temperature (175 and 183 °C) where head-in-pillow defects
but this was not necessary in the present evaluation. could occur, the warpage at the maximum tin-lead solder
profile (220 °C) and the maximum lead-free solder profile
Delamination measurements were compared before and (245 °C).
after reballing. A delamination area (or length) change was
defined as a percentage change computed from the 250

delaminated area (or length) divided by the total area (or 225
length) of interest. The following thresholds were used for 200
acceptance:
175

Temperature (°C)
o Any external crack visible using a 40X optical 150

microscope was evaluated to ensure that it did not 125

increase by more than 10%. 100

o Any internal cracks that intersected a bond wire, ball 75
bond or wedge bond were evaluated to ensure that
50
they did not increase by more than 10%.
o Any internal cracks extending from any internal 25

feature to the outside of the package were evaluated to 0

0 2 4 6 8 10 12 14
ensure that they did not increase by more than 10%. Time (minutes)
o No delamination on the active side of the die was
permitted. Fig. 5: Heating profile use for the warpage measurements
o If applicable, no delamination change > 10% on any (red circles indicate temperatures where warpage data was
wire bonding surface of the die paddle (down bond obtained)
area) or the lead frame of LOC (lead on chip) devices.
o If applicable, no delamination change > 10% along Postive
Negative Negative
Positive
any polymeric film bridging any metallic features that
are designed to be isolated (verifiable by through
transmission acoustic microscopy and/or CSAM
imaging from the opposite side as required). Fig. 6: Definition for positive and negative warpage
directions
For example, if the total die paddle area was 0.100 square
inches (62.5 square mm) and had a pre-existing Since warpage was measured through the heating and the
delamination of 3 percent (0.003 square inches or 1.88 cooling ramps, it was possible to verify that the BGA under
square mm) before reballing, the delamination would be test returned to its original warpage state after exposure to
allowed to increase to a value less than 13 percent of the the peak reflow temperature.
area of interest (0.013 square inches or 8.39 square mm)
The total component warpage was defined per JEITA ED-
after reballing.
7306, in which the measurement zone included only the
Warpage Measurement Method solder ball attachment area. Plots of warpage versus
Component warpage was measured according to JEDEC temperature were generated by averaging the data across
JESD22-B112 and JEITA ED-7306 standards. An each diagonal. Positive warpage was defined as the
Akrometrix TherMoiré PS200 system was employed, using condition where the center of the component was lifted from
a 100 line per inch grating. This grating provides a the seating plane, while negative warpage was defined as
measurement resolution of at least 2.54 microns (0.1 mils), the condition where the corners of the component were
of out-of-plane displacement. lifted, as indicated in Fig. 6.
 As originally published in the SMTA Preedings.

Coefficient of Thermal Expansion Measurement Method Shear and Ball Pull Test Method
All materials have a tendency to change in length in Solder ball pull and shear testing gives an indication of the
response to a change in temperature. This material property interconnect strength and failure mode in the ball attach pad
is called the material’s coefficient of thermal expansion region under various rates of loading. For this testing,
(CTE). Coefficients of thermal expansion for electronic comparisons were made between the as-received BGA
packaging materials can range from 4-ppm/°C for the devices and those that were reballed. Shear testing was
silicon die, to 17-ppm/°C for glass epoxy PWB material, to performed according to JEDEC JESD22-B117A, while pull
28.3-ppm/°C for tin-lead solder. The average CTE of a testing was performed according to JEDEC JESD22-B115.
BGA package varies with the materials and geometries used The failure force, energy and failure mode were recorded
in its construction. As with warpage, a change in the BGA for each trial and tabulated for each component.
material properties or internal delamination from the
reballing process might be revealed in differences in CTE. Testing speeds were selected based on the desired failure
mode. The goal of this testing was to determine the
robustness of the solder ball interface with the BGA
attachment pad and observe any changes due to the reballing
process. A test speed of 1-m/s was selected for shear test,
and 100-mm/s for pull test. These test speeds were selected
because they were expected to produce interfacial failures
(B) on lead-free solder joints based on previous testing
experience. Two BGAs of each type were selected for pull
(A) testing, and two additional BGAs were selected for shear
testing. Pull testing was performed on 16 solder balls per
Fig. 7: Typical section where plane along which the CTE is component, or 32 solder balls per component type.
measured. (L256 shown).
Sample preparation for shear testing required that the
In the current evaluation, Moiré interferometry [11] was component be depopulated of solder balls except for a single
used to measure the BGA CTEs at Binghamton University. outer row to be tested. For this reason, as well as component
Moiré interferometry is an optical method, providing clamping requirements, only 9 solder balls per M90 BGA
wholefield contour maps of in-plane displacements. In this and 12 solder balls per L256 BGA could be tested in shear.
method, a high frequency crossed-line diffraction grating is
replicated on the surface of the specimen and it deforms BGA Analytical Modeling Method
together with the underlying specimen. Coherent beams Finite-element analysis (FEA) was used to determine the
from a laser are used to create a virtual reference grating. internal stresses resulting from reballing. The analytical
The deformed specimen grating and reference grating effort concentrated on the ball removal process. During ball
interact to produce the Moiré fringe pattern with each fringe removal with liquid solder contacting the bottom of the
corresponding to the magnitude of movement between the BGA, the thermal conditions are considerably different than
BGA and the reference grid. with the surface mount convection reflow typically used for
soldering. In the present work, an overall warpage model
In the present evaluation, the CTE was measured on four that could be correlated to the warpage testing was created
BGA samples, two in an as-received condition and two that for the L256 and the F473 BGAs. In addition, the maximum
had been reballed. To measure the CTE of the BGAs, the L256 BGA package stresses during ball removal were
BGAs were cut along their diagonal as shown in Fig. 7. determined for two different ball removal thermal profiles.
Then a polymer grating was bonded to the BGA at a One profile was based on the measured package
temperature of 80 °C. Both the BGA and the grating temperatures during ball removal in the present work, and
contract upon cooling to room temperature and when the the other had a higher preheat and higher wave solder
virtual laser reference grid is projected onto the grating, temperature that might occur in a modified process. The
Moiré fringes are formed. For the present system sensitivity, maximum ball removal stresses were compared to the
each fringe represents a displacement of 0.417 microns and maximum standard surface mount reflow stress.
the CTE can be determined from the following computation:
First a finite-element half-symmetry model was developed
CTE = Coefficient of thermal expansion = (ΔL/L)*(1/ΔT) (see Fig. 8) and a transient thermal analysis was used to
L = Length of the BGA over the region of interest. determine the temperature distribution inside the part. The
ΔL = Change in length of the specimen computed thermal response was compared to the measured
with ΔL = Nf * Sensitivity of the equipment, which is temperatures at the front of the package, the top center of
0.417 micron displacement per fringe the die and the rear of the package during ball removal.
Nf = Number of fringes
ΔT = Change in temperature, °C
For example ΔT = 59.3 °C is the temperature difference
obtained when the replication temperature is 80°C and
room temperature is 20.7 °C.
 As originally published in the SMTA Preedings.

that the assembly was soldered. The prognostics circuitry

stores temperature data during the thermal cycling. The
boards utilize boundary scan circuitry and/or custom circuits
to monitor the continuity of select BGA balls during thermal
cycling. The assembled boards were placed in a thermal
cycling chamber and the chamber ambient was cycled from
-55 to +95 °C with half hour ramps and dwells. Since the
modules were widely spaced and exposed directly to the
circulating chamber air, their temperature had no more than
Fig. 8: BGA Finite-Element Model a ten minute lag from the chamber temperature during the
transitions.
When the thermal FEA was completed, the thermal
elements were converted to structural elements and a static BGA REBALLING PROCESS
structural analysis was conducted to calculate internal Reballing was performed using soldering methods in
stresses. The parametric modeling capabilities of the accordance with J-STD-001 Class 3. The balls were
ANSYS finite-element code was used to automatically removed using a flowing solder wave methodology with the
generate the model based on datasheet and measured part devices preheated to a range of 120 to 160 °C before
dimensions. The L256 BGA model has 54,921 nodes and contacting the tin-lead eutectic solder wave. The liquid
11,560 hexahedron (20-node) elements while the F473 BGA solder wave was maintained at 230 °C. A typical ball
model is larger with 97,008 nodes and 21,294 hexahedron removal thermal profile is shown in Fig. 10. The process
(20-node) elements. begins by placing the part in a fixture with the ball side
facing downward toward the wave. The fixture is then
Material properties for most inorganic electronic packaging placed on the moving machine fingers at a velocity of 80-
materials (excluding solders) are well defined and not cm/minute, moving through a preheating stage, the flowing
subject to large variations over typical electronic packaging solder wave, and finally to a supplier proprietary pad
temperature ranges, because typical service operation is at leveling process.
relatively low homologous temperatures. This assertion is
not valid for polymer materials when the glass transition The part is in contact with the flowing solder wave for
temperature (Tg) is exceeded. Thermal expansion data is approximately 5 seconds. The thick lines in Fig. 10 show
frequently used to determine the Tg, so thermal expansion the computed finite element model thermal responses of the
values above and below it are readily available as published package (front, die top center and rear) and the thin lines
material properties. However, the elastic modulus is also represent the measured data. Note that the BGAs cooled as
reduced above Tg, and it is important that models also it traversed the short distance between the preheater and the
include this effect. This relationship is especially significant solder wave, and again between the solder wave and the pad
for die bond materials, which usually have Tg values much leveling process. Throughout the reballing process, the
lower than typical molding compounds. Since soldering BGAs were treated as moisture sensitive devices in
process temperatures routinely exceed typical Tg values, accordance with J-STD-033. (Note: The CSAM evaluation
modeling of electronic package stresses during soldering parts were baked dry prior to reballing.)
requires that both thermal coefficient of expansion and
modulus changes above the Tg be included. Details on the After ball removal, the packages were cleaned to remove
temperature-dependent material modeling approach used ball removal flux and inspected. The ball attach proceeds
here can be found in Reference [13]. The specific material next. Flux was applied to the package, Sn63Pb37 balls were
properties used in the present work were calibrated so that placed on the pads, the package was reflowed in a standard
the steady-state thermal analysis matched the measured surface mount convection reflow oven (peak part
warpage and CTE values obtained during the testing. temperature = 225 °C), cleaned, and then inspected. The
Following calibration, the transient thermomechanical solder ball composition was in accordance with J-STD-006,
analysis was performed that simulated the BGA package Table A-2. The flux was a low activity flux (ORLO per J-
bottom moving over the flowing solder wave. STD-004).

Thermal Cycling Test Method After the alteration, the ball geometry was verified to insure
The reballed BGAs were soldered onto custom designed that it met the dimensional requirements of the original part
circuit boards to evaluate assembly level thermal cycling drawing, and inspected for any defects that would degrade
reliability. A typical test module assembly is shown in Fig. the operation or reliability. Component cleanliness was
9. The input/output connectors, interface circuitry and verified through ionic cleanliness testing and visual
prognostics health monitoring circuitry are on the left side inspection. After reballing, the parts were marked with a
of the card. There are 12 electrically monitored BGAs in the yellow dot to allow differentiation from non-reballed parts
center and the four BGAs for cross-sectioning are located on and serialized to facilitate tracking through testing.
the right break-off portion of the assembly. The modules are
individually serialized and marked with the year and month
 As originally published in the SMTA Preedings.

Front Center Rear

230

190

Temperature (deg C)
150

110

70

30

-10
0 50 100 150 200
Time (sec)

Fig. 10: Typical BGA ball removal thermal profiles (L256

shown). Fine lines are measured data and bold lines are the
model results. The package front, the die top center and the
package rear temperatures were measured and modeled.

Physical Measurements of BGA Features

The results of the overall measurements indicate that the
physical BGA package dimensions were not impacted to
any significant degree by the reballing process. The cross-
section and radiographic inspections did not reveal any
detectable changes or damage to the BGA packages with
respect to internal layers, bump configuration, inner layer
traces, wire bonding, die attachment, or passive components
where present.

A typical cross-section of the overall package is shown in

Fig. 13. In general, all BGA interconnects were constructed
of copper base foils with electroless nickel (e.g. NiP alloy) /
immersion gold (ENIG) finish on bump attach pads.All pads
were solder mask defined (SMD), with ENIG processing
occurring after the application of solder mask. Intermetallic
formations seen on the reballed BGAs all appear to be
robust with no indications of voiding, separation, excess
oxidation, porosity, or lack of solderability.

No internal damage was identified during the sectioning that

was not externally visible during the in-process inspection.
Cross-sections of the internal BGA interconnect circuit
boards were found to be acceptable to IPC-6012B and IPC-
A-600 for both the as-received and reballed BGAs. Sections
Fig. 9: Photograph of a typical thermal cycling test board were reviewed in both un-etched and chemically etched
(top), board chassis used in thermal chamber (middle) and preparations. Cross-sections were reviewed for solder mask
test equipment (bottom). coverage and registration, surface and inner layer copper
integrity, ENIG finish homogeneity, laminate integrity, and
RESULTS AND DISCUSSION plating process. No issues were detected in either the as
In-process Inspection received or reballed packages. A summary of the
All the packages passed the ball size dimensional measurements obtained from the cross-sections are given in
inspection. In addition, the BGAs were visually inspected Table 3. The BGA ball attach pad and solder ball metallurgy
after ball removal and again after ball re-attachment. Out of details of both the original lead-free ball and the reballed
the 360 BGAs processed in the present activity, only two tin-lead ball are shown in Fig. 14.
parts were rejected. One BGA had a missing pad and a
cracked pad (See Fig. 11) and one exhibited bottom side
delamination (Fig. 12).
 As originally published in the SMTA Preedings.

Typical photomicrographs of the electroless nickel layer

before and after reballing are shown in Fig. 15. As indicated
in Table 4, the reduction in average electroless nickel was
typically less than a micron after reballing. This minimal
level of dissolution is generally consistent with other
investigator’s findings that negligible ENIG dissolution
occurs after tin-lead reballing [12]. At a thickness of 4.34
microns, the X1148 had the thinnest average electroless
nickel, while at 8.84 microns the L256 had the thickest
nickel layer. The fact that the X1148 sample after reballing
had a slightly greater nickel thickness than the as-received
sample indicates that there is some part-to-part variation in
nickel layer thickness, probably less than a micron.

SAM and Visual Delamination Assessment Results

CSAM measurements and visual inspections at 40x were
performed on a minimum of 10 components in the “as-
received” condition and same components after reballing. A
sample CSAM image and typical inspection summary are
provided in Fig. 16. No delamination conditions exceeding
our acceptance criteria (defined in the Method section) were
found on the M90, L256 or the F473 BGAs. Note that the
X1148 could not be imaged because of the presence of a
cavity under the lid over the die. The M90 device exhibited
Fig. 11: Photomicrograph BGA X1148 SN1 package one suspect area that was verified by cross-sectioning not to
bottom after ball removal showing a missing pad (top) in be delamination (see Fig. 17).
one location and fracture around a pad in another location
(bottom). Anomalies were found after ball removal. Warpage Measurement Results
In general, BGA warpage could impact soldering processes
or increase solder stresses during thermal cycling in service.
In Fig. 18 through Fig. 21, the warpage results are given for
the as-received BGAs and the reballed BGAs.

Three of the BGA types (M90, L256 and F473) exhibited a

slight increase in warpage on the reballed samples. The
L256 BGA exhibited the greatest increase in warpage (50
microns at 245 °C). An increase in warpage after reballing
suggests that the polymer insulating materials have
increased in modulus (e.g. increased degree of cure).

The surface contours at 245 °C of the as-received and the

reballed L256 BGAs are shown in Fig. 22 and Fig. 23.
Examining the warpage of the L256 BGA further, Fig. 24
shows the surface contours at the typical maximum tin-lead
soldering temperature of 220 °C. At this temperature, the
warpage was 81 microns which compares well with the
warpage of the as-received tin-lead version of the L256
shown in Fig. 25.

Fig. 12: Photomicrographs showing BGA F473 SN13

solder mask crack (top) and ball side copper plane
separation from the BGA substrate interconnect (bottom)
found after ball attachment.
 As originally published in the SMTA Preedings.

Table 3: BGA Cross-section measurements (microns)

M90 L256 F473 X1148
Mold thickness
400.1 730.3 831.9 NA
adjacent to die
Mold thickness
222.3 575.3 501.7 NA
over die
Die thickness 143.5 162.6 279.4 793.8
Die adhesive
31.8 19.1 31.8 76.2
thickness
Interconnect
substrate 152.4 254 212.1 1073.2
thickness
Pad Cu
25.4 25.4 25.4 25.4
thickness
SMD pad
393.7 335.3 387.4 406.4
Fig. 13: Typical overall package cross-section (L256 lead- diameter
free BGA with the molding compound thicknesses shown) Solder mask
43.2 38.1 39.4 44.5
thickness
Table 4: Electroless nickel thickness (microns)
Avg Min Max Range
As-received
M90 6.87 6.53 7.21 0.69
L256 8.84 8.56 9.12 0.56
F473 6.67 6.40 6.93 0.53
Fig. 14: Typical solder mask defined BGA solder ball and X1148 4.34 4.06 4.62 0.56
pad BGA solder mask defined pad and ball metallurgy. Reballed
original lead-free ball (Left) and after reballing (Right). M90 5.72 5.59 5.84 0.25
(L256 lead-free BGA shown) L256 7.95 7.47 8.43 0.97
F473 5.52 4.78 6.27 1.50
X1148 4.43 4.37 4.50 0.13

Fig. 15: Typical photomicrographs of the electroless nickel

layer thicknesses. As-received part (top) and a part after
reballing (bottom). (L256 lead-free BGA shown)
Fig. 16: Typical CSAM image (top) and inspection
summary (bottom). (The F473 BGA shown)
 As originally published in the SMTA Preedings.

L256
150

125

100

75

50

Warpage (microns)
25

-25

-50

-75
#1 As Received
-100 #5 As Recevied
#4 Reballed
-125 #6 Reballed

-150
25 100 125 150 175 183 200 220 230 245 230 220 200 183 175 150 100 25
Temperature (°C)

Fig. 19: Warpage on as-received and reballed L256 BGAs

F473
150

125

100
Fig. 17: CSAM and cross-section images of a M90 BGA. 75
The CSAM images of the as-received (top left) and the
Warpage (microns)

50
reballed parts (top right) are shown. The red lines indicated 25
regions of suspected delamination. The cross-section 0
(bottom) shows that no delamination was present. -25

-50

-75 #1 As Received
#7 As Recevied
-100 #6 Reballed
M90 #8 Reballed
-125
150
-150
125
25 100 125 150 175 183 200 220 230 245 230 220 200 183 175 150 100 25
100
Temperature (°C)
75

50 Fig. 20: Warpage on as-received and reballed F473 BGAs

Warpage (microns)

25

-25

-50
X1148
-75 #1 As Received
#5 As Recevied 150
-100 #6 Reballed
#8 Reballed 125
-125
100
-150
25 100 125 150 175 183 200 220 230 245 230 220 200 183 175 150 100 25 75
Temperature (°C)
50
Warpage (microns)

Fig. 18: Warpage on as-received and reballed M90 BGAs 25

-25

-50

-75
#27 As Received
-100 #31 As Recevied
#21Reballed
-125 #22 Reballed
-150
25 100 125 150 175 183 200 220 230 245 230 220 200 183 175 150 100 25
Temperature (°C)

Fig. 21: Warpage on as-received and reballed X1148 BGAs

 As originally published in the SMTA Preedings.

Warpage values around 75 microns typically have not

resulted in soldering issues. The M90 and the F473 BGAs
reballed samples exhibited slightly more warpage than the
as-received parts. The warpage behavior trends of the
X1148 BGA were not clear and may have been complicated
by the large metal lid on the device.

Coefficient of Thermal Expansion Measurement Results

BGAs have two distinct CTE values. One is the CTE under
the die region, which is primarily driven by the die, because
of its low coefficient of expansion and high modulus. CTE
values under the die typically range from 6 to 12-ppm/°C.
Fig. 22: Warpage of as-received L256 SN1 at 245°C. The other CTE value of concern is the CTE from outer ball
Coplanarity = 94 microns to outer ball. Typically, the die dimension is a small portion
of the overall size of the BGA and the CTE from ball-to-ball
is driven by the BGA interconnect and overmold materials.
CTE values from ball-to-ball typically range from 9 to 15-
ppm/°C. Some typical Moiré fringe patterns are shown in
Fig. 26 and Fig. 27. A summary of the CTE data of as-
received and reballed BGAs is shown in Fig. 28. The CTE
of the reballed BGAs might be slightly reduced for the M90
BGA as compared to the as-received BGAs. However, the
CTEs for the remaining BGAs were similar between the as-
received and the reballed BGAs.

Fig. 23: Warpage of the reballed L256 SN6 at 245°C.

Coplanarity = 144 microns.

Replication Temperature = 80 °C and

Room Temperature = 22.5 °C then ΔT = 57.5 °C
Sample Length: 24.02mm and length of die: 5.84mm
CTE (End to End, Nf = 31) = 9.35ppm/°C
CTE (Under the die, Nf = 7) = 8.69ppm/°C
(fringe count was made approximately 0.5 mm below the
ball to package interface)
Fig. 26: L256 SN 16 Moiré fringe pattern for an as-received
condition BGA
Fig. 24: Warpage of the reballed L256 SN6 at 220 °C.
Coplanarity = 81 microns.

Replication Temperature = 80 °C and

Room Temperature = 20.7 °C then ΔT = 59.3 °C
Sample Length: 24mm and Die Length: 5.77mm
CTE (End to End, Nf = 33) = 9.66ppm/°C
CTE (Under the die, Nf = 6) = 7.31ppm/°C
Fig. 27: L256 SN 6 Moiré fringe pattern for a reballed BGA

Fig. 25: Warpage of the as-received tin-lead version of the

L256 SN5 at 220 °C. Coplanarity = 71 microns.
 As originally published in the SMTA Preedings.

BGA Ball Shear and Ball Pull Measurement Results High speed testing caused some brittle intermetallic and pad
Solder ball pull and shear testing gives an indication of the cratering failures in the lead-free balls. The results for the
interconnect strength and failure mode under various rates high speed ball pull and shear tests are shown in Fig. 29
of loading. The goal of this testing was to determine the through Fig. 36. The tests on the reballed tin-lead
robustness of the solder ball interface with the attachment components produced almost exclusively ductile failures
pad, and observe any changes as a function of the reballing (i.e. failure in the bulk solder), indicating a good bond
process. BGAs typically have three failure modes. between the solder and the metal pad to which it was
attached, and a good bond between the pad and the BGA
 Bulk solder failure interconnect. The “as-received” lead-free BGAs exhibited
o Failures in the bulk solder are ductile failures higher pull and shear forces as well as more instances of
typically observed in tin-lead assemblies or lead-free brittle (interfacial) failures than the reballed tin-lead
assemblies at low loading velocities. components. This result was expected because lead-free
 Intermetallic layer failure solders are stronger and stiffer than tin-lead solder, as
o Failures indicate a weak intermetallic layer or can be observed by other investigators [7].
due to the increased stiffness in the lead-free solder
balls. Both the L256 and the X1148 devices had samples that were
 Pad cratering failure tin-lead in as-received condition, which allowed for a direct
o Failure represents a strong intermetallic layer and comparison to the reballed devices. The results indicate that
more brittle PWB dielectrics. Typically observed in at worst, the reballed devices have the same strength, and at
lead-free assemblies. best, the reballed devices are 10% stronger than the as-
 Extrusion failure received tin-lead devices. The results provide confidence
o Occurs when the solder pulls out of the instrument that the reballing process results in sufficiently well attached
ball pull gripping tool. balls. In addition, since no intermetallic failures or pad
 Mixed mode failure cratering failures were observed, the reballing process
o Failure that exhibits a combination of intermetallic yielded sufficient interfacial strength.
and bulk solder failure that represents a pad that has
some weaker intermetallic regions.

16
end-to-end under die
14

12

10
CTE (ppm/oC)

0
#16 X1148 As Recd

#17 X1148 As Recd

#24 X1148 Reballed

#30 X1148 Reballed

#5 L256 Reballed

#6 L256 Reballed
#7 M90 Reballed

#8 M90 Reballed

#10 F473 Reballed

#14 F473 Reballed

#3 M90 As Recd

#4 M90 As Recd

#15 L256 As Recd

#16 L256 As Recd

#3 F473 As Recd

#4 F473 As Recd

Fig. 28: Summary of BGA CTE data of as-received and reballed BGAs
 As originally published in the SMTA Preedings.

M90 - Ball Shear X1148 - Ball Pull

As-Received Pb-Free As-Received SnPb
Force (grams) Energy (mJ) Failure Modes Force (grams) Energy (mJ) Failure Modes
Average 1444.9 2.1 4 Bulk Solders Average 2076.2 5.5 31 Bulk Solders
Minimum 1118.4 0.7 9 Interfacial 1956.6 4.3 0 Interfacial
Minimum
Maximum 1722.7 3.9 0 Pad
Maximum 2211.8 7.8 0 Pad
Range 604.3 3.2 0 Extrusions
Range 255.2 3.5 1 Extrusions
Std. Dev. 173.9 1.1 5 Mixed
Post-Reballing SnPb
Std. Dev. 67.9 0.7 0 Mixed
Force (grams) Energy (mJ) Failure Modes
Post-Reballing SnPb
Average 1129.8 2.4 18 Bulk Solders Force (grams) Energy (mJ) Failure Modes
Minimum 1041.5 1.9 0 Interfacial Average 2204.3 8.1 27 Bulk Solders
Maximum 1244.7 3.0 0 Pad Minimum 2077.0 4.8 0 Interfacial
Range 203.2 1.1 0 Extrusions Maximum 2328.3 10.8 0 Pad
Std. Dev. 55.1 0.3 0 Mixed Range 251.3 6.0 5 Extrusions
Fig. 29: M90 Ball shear results for as-received and reballed Std. Dev. 63.9 1.5 0 Mixed
BGAs Fig. 32: X1148 Ball pull results for as-received and reballed
BGAs
L256 - Ball Shear
As-Received SnPb
Force (grams) Energy (mJ) Failure Mode M90 - Ball Pull
Average 1076.7 2.2 24 Bulk Solders As-Received Pb-Free
Minimum 937.4 1.7 0 Interfacial Force (grams) Energy (mJ) Failure Modes
Maximum 1175.2 3.5 0 Pad Average 1202.2 2.3 20 Bulk Solders
Range 237.8 1.8 0 Extrusions Minimum 1063.3 1.2 5 Interfacial
Std. Dev. 45.1 0.4 0 Mixed Maximum 1274.0 3.5 0 Pad
As-Received Pb-Free Range 210.7 2.3 5 Extrusions
Force (grams) Energy (mJ) Failure Mode
Std. Dev. 54.0 0.5 1 Mixed
Average 1777.0 4.1 15 Bulk Solders
Post-Reballing SnPb
Minimum 1360.2 1.8 1 Interfacial
Force (grams) Energy (mJ) Failure Modes
Maximum 2203.6 5.6 0 Pad
Average 999.0 2.1 29 Bulk Solders
Range 843.4 3.9 0 Extrusions
Minimum 942.2 1.7 0 Interfacial
Std. Dev. 210.1 0.9 8 Mixed
Post-Reballing SnPb Maximum 1057.5 3.0 0 Pad
Force (grams) Energy (mJ) Failure Mode Range 115.3 1.3 2 Extrusions
Average 1221.8 2.4 24 Bulk Solders Std. Dev. 32.6 0.3 0 Mixed
Minimum 1107.4 2.1 0 Interfacial Fig. 33: M90 Ball pull results for as-received and reballed
Maximum 1328.3 3.3 0 Pad BGAs
Range 220.9 1.2 0 Extrusions
Std. Dev. 46.3 0.3 0 Mixed
Fig. 30: L256 Ball shear results for as-received and reballed
L256 - Ball Pull
BGAs As-Received SnPb
Force (grams) Energy (mJ) Failure Mode
F473 - Ball Shear
Average 1283.7 2.1 18 Bulk Solders
As-Received Pb-Free
Minimum 1109.0 1.3 11 Interfacial
Force (grams) Energy (mJ) Failure Mode
Maximum 1405.7 4.0 0 Pad
Average 1650.6 4.3 26 Bulk Solders
Range 296.7 2.6 0 Extrusions
Minimum 1441.6 3.1 0 Interfacial
Std. Dev. 77.2 0.6 3 Mixed
Maximum 1872.1 13.2 0 Pad
As-Received Pb-Free
Range 430.5 10.1 0 Extrusion
Std. Dev. 99.6 1.7 5 Mixed Force (grams) Energy (mJ) Failure Mode
Post-Reballing SnPb Average 1903.9 3.5 18 Bulk Solders
Force (grams) Energy (mJ) Failure Mode Minimum 1587.2 0.7 11 Interfacial
Average 1203.5 2.5 32 Bulk Solders Maximum 2125.1 6.4 1 Pad
Minimum 1048.7 2.1 0 Interfacial Range 537.9 5.6 2 Extrusions
Maximum 1292.4 2.8 0 Pad Std. Dev. 113.9 1.3 0 Mixed
Range 243.7 0.7 0 Extrusions Post-Reballing SnPb
Std. Dev. 60.3 0.2 0 Mixed Force (grams) Energy (mJ) Failure Mode
Fig. 31: F473 Ball shear results for as-received and reballed Average 1535.8 2.8 31 Bulk Solders
Minimum 1423.0 2.2 0 Interfacial
BGAs Maximum 1641.5 6.1 0 Pad
Range 218.5 3.9 0 Extrusions
Std. Dev. 55.7 0.7 1 Mixed
Fig. 34: L256 Ball pull results for as-received and reballed
BGAs
 As originally published in the SMTA Preedings.

F473 - Ball Pull Pb-Free #1 Pb-Free #5 Model (94% filled, 100C Tg)
As-Received Pb-Free 150
Force (grams) Energy (mJ) Failure Mode
Average 1193.3 1.4 16 Bulk Solders 100
Minimum 1063.0 0.6 0 Interfacial
Maximum 1256.0 3.0 14 Pad
50
Range 193.0 2.4 1 Extrusion

Warpage (micron)
Std. Dev. 39.8 0.6 0 Mixed
Post-Reballing SnPb 0

Force (grams) Energy (mJ) Failure Mode

Average 1038.6 1.6 26 Bulk Solders -50
Minimum 870.5 0.9 0 Interfacial
Maximum 1181.2 2.7 0 Pad -100
Range 310.7 1.7 6 Extrusions
Std. Dev. 55.9 0.5 0 Mixed
-150
Fig. 35: F473 Ball pull results for as-received and reballed 0 50 100 150 200 250
BGAs Temperature (deg C)

Pb-Free #1 Pb-Free #7

X1148 - Ball Shear Model (95% filled, 100C Tg) Model (94% filled, 100C Tg)
150
As-Received SnPb
Force (grams) Energy (mJ) Failure Modes
100
Average 1825.3 4.0 30 Bulk Solders
Minimum 1664.1 3.2 0 Interfacial
Maximum 1929.0 5.0 0 Pad 50

Warpage (micron)
Range 264.9 1.8 0 Extrusions
Std. Dev. 73.0 0.5 0 Mixed 0
Post-Reballing SnPb
Force (grams) Energy (mJ) Failure Modes -50
Average 1888.7 5.1 32 Bulk Solders
Minimum 1719.3 3.7 0 Interfacial
-100
Maximum 2099.4 6.7 0 Pad
Range 380.1 3.0 0 Extrusions
-150
Std. Dev. 82.8 0.6 0 Mixed
0 50 100 150 200 250
Fig. 36: X1148 Ball shear results for as-received and Temperature (deg C)

reballed BGAs Fig. 37: Comparison between computed and measured

warpage. for the L256 (A) and F473 (bottom). Square and
round symbols represent two individual sets of measured
warpage data for each BGA type.
BGA Analytical Modeling Results
The transient thermal modeling results for the L256 Front Rear Center
correlated well with the measured temperatures during ball 250
removal as was previously illustrated by the bold lines in
Fig. 10. The maximum temperature of the die (top center) 200
was 216.5 °C during ball removal.
Temperature (deg C)

150
Next the static warpages of the BGAs were evaluated. The
BGA modeling material properties were selected to match 100
the static warpage measurements. Fig. 37 shows that the
analytical prediction for overall warpage correlated well to 50
the measured data for the L256 and the F473 BGAs.
0
0 50 100 150 200 250 300 350 400 450
Time (sec)

Fig. 38: Thermal analysis results of the package temperature

during a typical surface mount reflow soldering process.
Temperatures for the front of the package, the top center of
the die and the rear of the package were computed to be
nearly the same.
 As originally published in the SMTA Preedings.

SMT Reflow Stress Actual Ball Removal Profile Stress The model reveals that there is interaction between the rate
Higher Temp. Ball Removal Stress SMT Reflow Die Temp. of temperature change and the maximum temperature that
Actual Ball Removal Profile Die Temp. Higher Temp. Ball Removal Die Temp.
influences the maximum package stress. As highlighted in
120% 240 Table 5, there are two instances where high stress occurs
100% 200
near the peak temperature. Furthermore, the higher the
package temperature, the higher the maximum stress. This
Maximum Relative Stress

Temperature (deg C)
80% 160 suggests that both the temperature gradient and the die (top
center) temperature contribute to the stress state. A detailed
60% 120
study of this behavior is the topic of future work.
40% 80
Table 5: Summary of computed peak temperatures and
20% 40
relative stresses
0% 0 Time Relative
0 1 2 3 4 5 6 7 Time from Die Temp. Mold
Time (min)
(sec) Tmax (°C) Stress
Fig. 39: Comparison of maximum relative package stresses SMT 270.4 -36.2 198.5 100.0%
for three different thermal profiles. The stresses are Reflow 306.6 0.0 216.3(max) 94.7%
normalized to the maximum stress observed during surface 332.6 26.0 199.5 99.9%
mount reflow soldering. 141.3 -6.9 208.8
Actual 87.8%
ball 148.2 0.0 216.5(max) 87.3%
removal 154.3 6.0 199.4 92.3%
Higher 60.9 -12.4 168.1 97.7%
temp. ball 73.3 0.0 228.9(max) 91.3%
removal 82.0 8.7 195.6 101.0%

Thermal Cycling Results

The M90, L256 and F473 test module assemblies are
currently being thermal cycled. The thermal cycling results
are summarized in Table 6. Each board has 12 BGAs that
are continuously electrically monitored using boundary scan
Fig. 40: Stress plot of L256 BGA during ball removal circuitry. The F473 module does not have operational
thermal exposure. The left image shows the solder wave monitoring circuitry and was cross-sectioned at 136 cycles
front when it is mid way in its traverse across the package at to assess the integrity of the solder joints. No assembly
2.26 minutes (135.59 seconds). The right image show the solder fractures or other anomalies were observed (See Fig.
stress plot when the maximum stress occurs at 2.57 minutes 41). Other investigators have observed good assembly
(154.26 seconds). The highest stresses are red shades and thermal cycling performance of reballed BGAs after 1000
low stresses are blue shades. cycles (-25° C/+125 °C, 15 minute ramps and dwells), as
well as vibration and shock [14].
The model was then used to compare the stresses in the
BGA during the ball removal. The stresses are normalized Table 6: Thermal cycling results (-55 to +95 °C ½ hour
to the maximum stress that would normally be observed ramps and dwells)
during surface mount reflow (stresses equal to the SMT BGA Module Number
reflow stress are plotted as a magnitude of 100%). The Type Serial of Comments
surface mount reflow thermal profile has a maximum Number Cycles
temperature of 216.3 °C at the die (top center), shown for No electrical failures
reference in Fig. 38. In contrast to the actual ball removal M90 M1 675
and cycling continuing
process (Fig. 10), there is minimal temperature variation
No electrical failures
across the package during the SMT reflow profile. The L256 L1 281
and cycling continuing
stresses computed during reballing are compared to the
Non-monitored: Cross-
SMT reflow stresses in Fig. 39. In addition, to assess the F473 F1 136
sectioning complete
actual ball removal profile, the stress for a ball removal
Non-monitored: Cross-
profile with a higher preheat temperature and a higher die
F473 F2 663 sections planned at
temperature was evaluated. The computed maximum stress
1000 cycles
during the actual ball removal thermal profile was less than
the surface mount solder reflow. The maximum stress was
located around the corners of the die as shown in Fig. 40.
The higher temperature ball removal profile resulted in the
largest stresses.
 As originally published in the SMTA Preedings.

temperatures). In addition, the modeling effort will be

extended to develop a basis for similarity assessments
among different BGA packages intended for reballing.

ACKNOWLEDGEMENTS
Thanks to Jim Carrigan, David Grant, and Hal Rotchadl of
Premier Semiconductor Inc. for reballing services, visual
inspection and thermal characterization and to Wayne Jones
and Brian Roggeman at Universal Instruments Corp., in
Conklin, NY, for the BGA characterization service. In
addition, the authors would like to thank Dr. Park at the
Binghamton University Mechanical Engineering
Optomechanics Laboratory in Binghamton, NY for his help
with the CTE measurements. The authors also wish to thank
Dr. Sam Saha, Ted Hartford and Joe Kane from BAE
Systems in Johnson City for their assistance on this project.

REFERENCES
[1] European Union (February 13, 2003). Directive
2002/95/EC/ of the European Parliament and of the
Fig. 41: Typical cross-section of F473 after 136 thermal
Council of 27 January 2003 on the restriction of the use
cycles from -55 to +95 °C.
of certain hazardous substances in electrical and
SUMMARY electronic equipment. Official Journal of the European
With careful attention to detail, BGA reballing remains a Union.
viable solution to manage the obsolescence of tin-lead ball
[2] Ganesan S, Pecht M. Lead-free Electronics, John Wiley
metallurgy. The cross-section evaluation provided the key
& Sons, Inc. NY, USA, 2006.
package construction details needed for modeling. The
scanning acoustic microscopic examination was generally [3] “The Lead Free Electronics Manhattan Project - Phase I
successful in assuring that the package structures did not Report,” Completed under U.S. Government Contract
delaminate during the reballing process, though it was not No. N00014-08-D-0758, The Benchmarking and Best
useful for the package with the lid. Visual inspection of the Practices Center of Excellence at ACI Technologies,
package is an important part of the reballing quality Inc., Philadelphia, PA, July, 30 2009
verification. Warpage measurement appears to be more http://www.navyb2pcoe.org/b2p_news.html and
useful than coefficient of thermal expansion measurement in https://acc.dau.mil/leadfree
assessing BGA changes during reballing. However, warpage
change acceptance limits are still need. The increase in [4] GEIA-HB-0005-2, Technical Guidelines for Aerospace
warpage observed after reballing suggests that the polymer and High Performance Electronic Systems Containing
insulating materials may have increased in modulus (e.g. Pb-free Solder and Finishes, Tech America Inc.,
increased degree of cure). The ball shear and ball pull tests Arlington, VA USA (formerly Government Electronic
verified that the pad-to-package interconnections were not Industries Association)
compromised by reballing. The thermomechanical stress
modeling effort needed the cross-sectioning, warpage and [5] Woodrow T.A., JCAA/JG-PP Lead-Free Solder Project:
CTE measurements. In general, these preliminary modeling -55 to +125 ° C Thermal Shock Test; March 1, 2006.
efforts suggest that increasing maximum temperature of the Also published in The Proceedings of SMTA
die (top center) during ball removal increases the maximum International Conference, Rosemont, IL, September 24-
stresses within the part and that the location of the 28, 2006.
maximum stress is near the die corners and edges. It is
hopeful that efforts such as this will contribute to the body [6] Woodrow T.A., The Effects of Trace Amounts of Lead
of knowledge needed for industry standard development on the Reliability of Six Lead-Free Solders,
[15]. Proceedings of the 3rd International Conference on
Lead-Free Components and Assemblies, San Jose, CA
FUTURE WORK April 23-24, 2003.
Assembly thermal cycling is continuing and assembly
vibration and mechanical shock testing are planned. The [7] Nie L, Osterman M, Pecht M, Song F, Lo J, Lee S W,
X1148 BGA module assemblies will be completed and Solder Ball Attachment Assessment of Reballed Plastic
subjected to thermal cycling, vibration and mechanical Ball Grid Array Packages, IEEE Trans. Components.
shock. Further modeling work will be pursued to assess and Packaging Tech., Vol. 32, No. 4, December 2009,
process sensitivity of ball removal variables (e.g. preheat pp. 901-908
temperatures, ball removal temperatures or ball attach
 As originally published in the SMTA Preedings.

[8] Cirimele R, BGA Reballing Reliability: A Study of

Multiple Reball Processes Looks at Copper Dissolution
and Functionality, Circuits Assembly, June 2009, pp.
27-29.

[9] Muonio J, and Stadem R, Solder Ball Attachment Using

Laser Soldering, Circuits Assembly, October 2008.

[10] Winslow R, BGA Reballing Overview, Presentation at

the Aerospace Industries Association – Lead-free
Electronics in Aerospace Project (AIA-LEAP) working
group meeting , June 3, 2007.

[11] Han B and Guo Y, Determination of an Effective

Coefficient of Thermal Expansion of Electronic
Packaging Components: A Whole-Field Approach,
IEEE Trans. on Components, Packaging and
Manufacturing Technology-Part A, Vol. 19, No. 2, June
1996, pp 240 - 247

[12] Cruz MM, and Winslow R, Interfacial Reactions

between Sn63-Pb37 Solder and Electroless/Electrolytic
Nickel UBM, University of Maryland CALCE Tin
Whisker and Lead-free Rework Workshop, College
Park Maryland, November 2008.

[13] Meschter S, Mckeown S, Effect of Hot Solder Dipping

on Part Stresses, Proceedings of IMECE2008, 2008
ASME International Mechanical Engineering Congress
and Exposition, October 31-November 6, 2008, Boston
MA.

[14] Davisson M, An OEM Perspective on Re-balling and

case history, University of Maryland CALCE Tin
Whisker and Lead-free Rework Workshop, College
Park Maryland, November 2008.

[15] GEIA-STD-0015, Requirements for BGA Reballing in

Aerospace and High Performance Electronics, Draft
Document, Tech America Inc., Arlington, VA USA
(formerly Government Electronic Industries
Association)